[RSSAC Caucus] Number of instances in the rss for a past date

[Wes Hardaker]

>
>
>
> While that may be true, I'm curious whether it's known why the load times
> vary across three orders of magnitude?
>

I can only offer speculations based on our own deployment architectures,
but it should serve as an example of why these measurements are somewhat
meaningless in the first place.

Consider a set of systems that are all on SOA-1.  Now, take one of those
systems out of the current processing load because it needs an OS update
while the rest continue handling the load.  A second later, the rest of the
systems upgrade to SOA-2.  A long time later the OS upgrade gets finished,
it gets rebooted and the name server comes on line and logs that it now has
SOA-2.  This could easily be order hours later given the length of an OS
upgrade and will affect the maximum time seen quite severely.  You can
likely start thinking of much more challenging scenarios like systems that
are entirely off for a hardware replacement, or network connectivity fix,
etc that could easily sneak into order days of delay in getting a new SOA.
This delay does not matter though, as none of these "down" systems would be
actually answering queries yet so there is no harm in the actual delay that
is being reported.

Many many more scenarios are likely as well, of course.  Routing outages
that cause no user degradations can technically cause reported load-time
slips, for example.
-- 
Wes Hardaker
USC/ISI
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mm.icann.org/pipermail/rssac-caucus/attachments/20231214/9b6d07b6/attachment.html>